Apple Has Silently Removed the UltraFusion Interconnect on M3 Max

When Apple revealed the existence of the M1 Ultra SoC back in 2021, it announced it had hidden an ultra-fast connector on the edge of the previously-announced Max chip. This allowed two Max chips to be joined to function as a single chip, resulting in the M1 Ultra. Apple repeated this pattern with the M2 Max and Ultra but has ditched it on the M3 Max, according to a new analysis. The switcheroo by Apple opens the floodgates for what is possible with the upcoming M3 Ultra SoC.

Word of Apple's potential plans for the M3 Ultra comes from Vadim Yuryev on X, who runs the YouTube channel MaxTech. He notes there is no UltraFusion interconnect on the M3 Max based on photos posted online and flagged by Wccftech, which raises many questions about the direction Apple is headed with its next Ultra SoC. According to one theory, removing the interconnect will let Apple go wild with its design, as it's no longer hobbled by efficiency considerations required for a notebook processor. That means no more efficiency cores, as the Ultra chip is only used in its Studio and Pro desktops.

The M3 Max sports 92 billion transistors, so the Ultra version could be more than double that number. Credit: Apple

This means Apple will be fashioning a custom piece of silicon for the next Ultra chip, and it will be gigantic. It will likely only feature performance and GPU cores, and even more tantalizing is it's possible it could have its own UltraFusion chip, allowing two of them to be combined for the long-rumored "Extreme" SoC that Apple reportedly cancelled previously. Its cancellation was supposedly because it would be too expensive to create for a niche audience of "Mac power users," which is our phrase, not Apple's.

The rumored M3 Ultra would scale better than linking two dies as there's no need to pass data through an interconnect, but it would also be a gigantic monolithic design. That will allow Apple to add as many performance and GPU cores as possible rather than being limited to just 2X of the Max SoC. This direction makes sense for a desktop-only chip, though. As always, the price has to be factored in here, as it sounds like it would be ungodly expensive.

We could see Apple go wild on an Ultra chip unrestrained by efficiency requirements for a notebook. Credit: Apple

The final piece of the puzzle here is that this chip might be so ambitious that it may not arrive until the M4 is launched, which will probably be later this year. That chip is expected to remain on TSMC's 3nm process but will use an enhanced version with improved performance and efficiency. TSMC isn't producing 2nm chips yet, which aren't expected to arrive until 2025.